This website uses Cookies. Click Accept to agree to our website's cookie use as described in our Privacy Policy. Click Preferences to customize your cookie settings.
Accept
Reject
Preferences
cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Choose Language Hide Translation Bar
burakbagdatli
burakbagdatli
Level III

Max Entropy Sampling favors the lower limit

Created: Aug 4, 2022 09:58 AM  |  Last Modified: Jun 8, 2023 05:10 PM (232 views)

When I create small maximum entropy designs, I get unbalanced distributions for the variables. It's mostly regular and with some extra points at the ends; however, the lower end is sampled much more frequently than any other point. I have tried it many times and attached a few examples to this message. If needed, I can provide more.

 

X1_Discrete.png

As you can see from the screenshot above, -1 is sampled must more frequently than any other value. Even with a continuous distribution it's clear that the lower limit is sampled more frequently:

X1_Continuous.png

However, it doesn't always happen. Unfortunately, in my experience, it happens more often than not. I have attached a few designs where the issue appears and disappears.

What we see of the real world is not the unvarnished real world but a model of the real world, constructed so it is useful for dealing with the real world. —Richard Dawkins
20vars_MaximumEntropy_200runs_Issue.jmp
20vars_MaximumEntropy_1000runs_NoIssue.jmp
13vars_MaximumEntropy_910runs_Issue.jmp
0 Kudos
1 REPLY 1
Mark_Bailey
Staff Mark_Bailey
Staff

Re: Max Entropy Sampling favors the lower limit

Aug 4, 2022 10:38 AM (228 views)  |  Posted in reply to message from burakbagdatli 08-04-2022

It is very difficult to assess the design one factor at a time. The solution must satisfy a multivariate criterion. Looking at two factors at a time, still short of all 20 factors, shows a smoother spread of runs.

 

matrix.PNG

 

If you believe that there is a problem with your design, you should reach out to JMP Technical Support (support@jmp.com) for resolution. If so, please reply with their findings and your conclusion.